inclusionAI/LLaDA2.2-flash

如何在Transformers中使用inclusionAI/LLaDA2.2-flash：

如何在vLLM中使用inclusionAI/LLaDA2.2-flash：

如何在SGLang中使用inclusionAI/LLaDA2.2-flash：

如何在Docker模型运行器中使用inclusionAI/LLaDA2.2-flash：

LLaDA2.2-flash是LLaDA2系列中的一个面向代理的扩散语言模型。通过将Levenshtein编辑（带有DELETE和INSERT控制标记）引入扩散语言建模，它代表了LLaDA2系列在代理应用中的第一步，包括长上下文工具使用、多轮交互以及稳健的错误修正。更多信息请参阅我们的技术报告：https://github.com/inclusionAI/LLaDA2.X/blob/main/LLaDA2_2_tech_report.pdf。

下表比较了LLaDA2.2-flash和Ling-2.6-flash在代理基准分数和吞吐量（TPS）方面的差异。

LLaDA2.2-flash评估设置：SWE-bench系列使用Claude Code框架进行评估。在所有基准测试中，我们使用128K上下文窗口，temperature=1.0，block_length=32，threshold=0.5，以及editing_threshold=0.0。每个分数代表五次运行的平均值。

† Ling-2.6-flash在SWE-bench Verified上的得分取自Ling和Ring 2.6技术报告，该得分使用OpenHands框架获得。Ling-2.6-flash在τ²-Bench、Claw-Eval和PinchBench上的得分也来源于技术报告，而其SWE-bench Pro和SWE-bench多语言得分是我们使用与LLaDA2.2-flash相同的Claude Code框架评估的。

Ling-2.6-flash评估设置：启用了MTP，使用4个草稿标记。

更多结果将在即将发布的技术报告中公布。

高效的128K扩散基础设施：LLaDA2.2-flash将上下文窗口扩展到128K，并引入Block Routing，将MoE专家激活限制在扩散块级别，以实现高效的长上下文代理工作负载。

Levenshtein编辑：我们引入了DELETE和INSERT控制标记，允许扩散解码编辑序列结构、删除冗余内容，并在并行生成期间创建插入槽。

能动强化学习：我们提出了基于Levenshtein编辑ELBO的块级策略优化（L-EBPO），利用能动环境奖励在多轮工具使用场景中训练Levenshtein编辑和错误修正。

LLaDA2.2-flash具有以下规格：

请确保已安装transformers及其依赖项。

为了实现最佳性能，我们建议使用以下设置作为起点：

采样参数：使用block_length=32，temperature=0.0，top_p=None，top_k=None作为稳定的默认设置。

去噪阈值：根据应用所需的速度-质量权衡调整threshold、editing_threshold和max_post_steps。较低的阈值可能提高推理速度，但可能导致重复增加或输出不稳定。

长上下文能动任务：对于长上下文工具使用和多轮能动应用，我们建议使用SGLang作为服务后端。请确保服务堆栈已为128K上下文窗口和模型的MoE扩散推理要求进行配置。

如果您在中国大陆，我们强烈建议通过🤖 ModelScope访问我们的模型：https://modelscope.cn/models/inclusionAI/LLaDA2.2-flash

SGLang部署支持即将推出。

本项目根据Apache许可证2.0条款授权：https://www.apache.org/licenses/LICENSE-2.0

如有问题、合作机会或反馈，请通过Hugging Face联系我们：https://huggingface.co/inclusionAI/LLaDA2.2-flash 或在仓库中提交issue：https://github.com/inclusionAI

加入我们，共同推动面向能动应用的开放、高效、智能的扩散语言模型的发展。

How to use inclusionAI/LLaDA2.2-flash with Transformers:

How to use inclusionAI/LLaDA2.2-flash with vLLM:

How to use inclusionAI/LLaDA2.2-flash with SGLang:

How to use inclusionAI/LLaDA2.2-flash with Docker Model Runner:

LLaDA2.2-flash is an agent-oriented diffusion language model in the LLaDA2 series. By introducing Levenshtein Editing (with DELETE and INSERT control tokens) to diffusion language modeling, it represents the LLaDA2 series' first step in agentic applications, including long-context tool use, multi-turn interaction, and robust error correction.For more information, please refer to our technical report：https://github.com/inclusionAI/LLaDA2.X/blob/main/LLaDA2_2_tech_report.pdf.

The following tables compare LLaDA2.2-flash and Ling-2.6-flash in terms of agentic benchmark scores and throughput (TPS).

LLaDA2.2-flash evaluation setup: The SWE-bench series was evaluated using the Claude Code scaffold. Across all benchmarks, we used a 128K context window with temperature=1.0 , block_length=32 , threshold=0.5 , and editing_threshold=0.0 . Each score represents the average of five runs.

† The Ling-2.6-flash score on SWE-bench Verified is taken from the Ling and Ring 2.6 Technical Report, where it was obtained using the OpenHands scaffold. The Ling-2.6-flash scores on τ²-Bench, Claw-Eval, and PinchBench are also sourced from the technical report, whereas its SWE-bench Pro and SWE-bench Multilingual scores were evaluated by us using the same Claude Code scaffold as LLaDA2.2-flash.

Ling-2.6-flash evaluation setup: MTP was enabled with 4 draft tokens.

More results will be released in the upcoming technical report.

Efficient 128K Diffusion Infrastructure : LLaDA2.2-flash extends the context window to 128K and introduces Block Routing , which bounds MoE expert activation at the diffusion-block level to enable efficient long-context agentic workloads.

Levenshtein Editing : We introduces DELETE and INSERT control tokens, allowing diffusion decoding to edit sequence structure, remove redundant content, and create insertion slots during parallel generation.

Agentic Reinforcement Learning : We propose Levenshtein Editing ELBO-based Block-level Policy Optimization (L-EBPO) , which leverages agentic environmental rewards to train levenshtein editing and error correction in multi-turn tool-use scenarios.

LLaDA2.2-flash has the following specifications:

Make sure you have transformers and its dependencies installed.

To achieve optimal performance, we recommend starting with the following settings:

Sampling Parameters : Use block_length=32 , temperature=0.0 , top_p=None , and top_k=None as stable default settings.

Denoising Thresholds : Tune threshold , editing_threshold , and max_post_steps according to the speed-quality trade-off required by the application. Lower thresholds may improve inference speed but can lead to increased repetition or unstable outputs.

Long-Context Agentic Workloads : For long-context tool-use and multi-turn agent applications, we recommend using SGLang as the serving backend. Please ensure that the serving stack is configured for the 128K context window and the model's MoE diffusion inference requirements.

If you are in mainland China, we strongly recommend accessing our model from 🤖 ModelScope：https://modelscope.cn/models/inclusionAI/LLaDA2.2-flash

SGLang deployment support is coming soon.

This project is licensed under the terms of the Apache License 2.0：https://www.apache.org/licenses/LICENSE-2.0.

For questions, collaboration opportunities, or feedback, please reach out via Hugging Face：https://huggingface.co/inclusionAI/LLaDA2.2-flash or open an issue in the repository：https://github.com/inclusionAI.

Join us in advancing open, efficient, and intelligent diffusion language models for agentic applications.